Lessons Learned from Multiobjective Automatic Optimizations of Classical Three-Site Rigid Water Models Using...

Area of Science:

• Computational chemistry
• Materials science
• Chemical physics

Background:

• Accurate water models are crucial for molecular simulations.
• Three-site rigid water models are widely used but have limitations.
• Current optimization methods focus on macroscopic properties, leaving room for improvement.

Purpose of the Study:

• To develop more accurate three-site rigid water models.
• To investigate the impact of incorporating microscopic data into model optimization.
• To enhance the predictive power of water models for molecular simulations.

Main Methods:

• Utilized Swarm-CG, a multiobjective particle-swarm-optimization algorithm.
• Trained water models using a combination of microscopic and macroscopic experimental observables.
• Focused on reproducing experimental radial distribution functions of liquid water at various temperatures.

Main Results:

• Models trained with microscopic data showed improved agreement with experimental observables.
• Incorporating microscopic information led to state-of-the-art accuracy efficiently.
• Analysis revealed the specific benefits of microscopic data in refining water models.

Conclusions:

• Optimizing water models with microscopic data is a highly effective strategy.
• This approach significantly enhances the accuracy of three-site rigid water models.
• The findings provide a framework for optimizing approximated molecular models in general.